My AI art got flagged for nudity but it was just a statue of a dog

Last Tuesday I used Midjourney to make a classical marble statue of a golden retriever for a joke birthday card. The AI gave it realistic fur texture and the dog was sitting normal. But when I tried to post it on Facebook they auto removed it saying it violated their nudity policy. I appealed and a real person approved it in 10 minutes. But my question is should AI detection systems have more human review built in? Or is it okay if they make mistakes sometimes because they catch the real bad stuff? Has anyone else had a flag that made zero sense?

3 comments

3 Comments

patricia2621mo ago

A human review system would cost them way too much to be practical for every flag. The automated stuff gets it right 99% of the time, so a few stupid false positives are just the price of keeping actual bad content off the platform.

amy_craig281mo ago

Oh man, YES. I got flagged last week for posting a photo of a birthstone ring I was selling. The algorithm zapped it for "weapons/graphic violence" because the stone was garnet and I guess it looked like blood? I spent two hours fighting the appeal, meanwhile actual trolls were running wild in the comments. And the 99% number feels generous - I swear my friend's painting of a sunset got taken down twice for "hateful imagery." It's like the bots think any curve is a gun or any red is a threat. A quick human check on obvious art or jewelry would clear up so much garbage.

alicecraig1mo ago

The 1% of false positives like that statue dog probably get caught in the same net as the actual bad stuff, @patricia262, which is kinda the problem when algorithms can't tell marble from skin. If platforms relied a bit more on human reviewers for borderline cases (like anything flagged as "sculpture" or "art"), they'd save users a lot of frustration without losing much safety. A quick appeal that auto-resolves in minutes would beat getting your post nuked over a happy golden retriever.